Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Attention!
This is a special internet edition of the article
The Avestan language and its problems
by Jost Gippert (2000).
It should not be quoted as such. For quotations, please refer to the original edition in
Nicholas Sims-Williams (ed.), Indo-Iranian Languages and Peoples,
Oxford: British Academy / Oxford University Press 2002
(Proceedings of the British Academy, 116), 165-187.
JOST GIPPERT
DURING THE LAST TWO DECADES of the twentieth century, Avestan studies
have witnessed an astonishing explosion of scholarly interest, which has been
manifested in a large number of monographs concerning both the Old Avestan
language, as represented in Zoroasters Gths and the Yasna Haptahiti,1 and
Young Avestan texts such as the Zamyd Yat or the rbadastn, each of which
has been edited twice within the past ten years.2
Comparing the competing editions now available, we may note that a high
level of agreement has been achieved both in the analysis of individual Avestan
word-forms and in the interpretation of the texts which depends on this analysis.
Nevertheless every scholar who deals with the language of Zoroaster will admit
that the Avestan tradition still conceals many a riddle to be solved by later
generations of investigators. In principle, this may be due to three well-known
reasons:
First we have to take into account the fact that the Avestan corpus is rather
small: less than 1,700,000 characters if arranged in a plain text format. If we com-
pare this with the Vedic tradition, which is what comes nearest to it with respect
both to its contents and to the circumstances of its transmission, we may observe
that the size of the complete Avestan corpus is less than that of the R gveda
Sahit alone, although the latter represents only one tenth of the whole Veda.
As for the Old Avestan parts, they are just half of the extent of the only other Old
1
Cf. J. Narten, Der Yasna Haptahiti, Wiesbaden 1986; J. Kellens / E. Pirart, Les textes vieil-
avestiques, 1-3, Wiesbaden 1988-1991; H. Humbach, The Gths of Zarathutra and the Other Old
Avestan Texts, 1-2, Heidelberg 1991 (hereafter: Gths).
2
Cf. Ph.G. Kreyenbroek, Sraoa in the Zoroastrian Tradition, Leiden 1985; A. Panaino, Titrya, 1,
Rome 1990; K. M. JamaspAsa (ed.), The Avesta Codex F1, Wiesbaden 1991; H. Humbach and J.
Elfenbein, rbedestn, Munich 1990; F. M. Kotwal and Ph. G. Kreyenbroek, The Hrbedestn and
Nrangestn, 1, Paris 1992; A. Hintze, Der Zamyd-Yat, Wiesbaden 1994; H. Humbach and P. P.
Ichaporia, Zamyad Yasht, Wiesbaden 1998. For recent works concerning the Avestan language and
its transmission in general, cf. J. Narten, Die Am a Sptas im Avesta, Wiesbaden 1982; J. Kellens,
Le verbe avestique, Wiesbaden 1984; R. S. P. Beekes, A Grammar of Gatha-Avestan, Leiden 1988;
K. Hoffmann and J. Narten, Der Sasanidische Archetypus, Wiesbaden 1989 (hereafter:
Sasan.Arch.); K. Hoffmann and B. Forssman, Avestische Laut- und Flexionslehre, Innsbruck 1996
(hereafter: ALF).
166 Jost Gippert
Iranian corpus available to us, namely, that of the Old Persian inscriptions.3
Under these circumstances it is not surprising that we are faced with an extremely
large number of hapax legomena which make many Old Avestan passages practi-
cally unintelligible even today.
The second reason why we cannot expect to find ad hoc solutions to the prob-
lems offered by Avestan tradition consists in the fact that we do not dispose of an
immediate descendant of the Avestan language in Middle and New Iranian times.
In this respect, too, Avestan is different from both Vedic and Old Persian. The
former was succeeded by other varieties of Sanskrit (Epic, Classical, Buddhist),
as well as by written and vernacular varieties of Middle Indic, which later
developed into the modern Indo-Aryan languages; as for Old Persian, we may
claim that, in spite of some temporary breaks in the tradition, its linguistic heri-
tage was by and large transmitted through Middle Persian into Modern Persian,
so that secondary evidence is available from these later stages of the language. It
is true, of course, that most of the Middle Persian texts we dispose of were inter-
related directly or indirectly with the Avestan corpus, in that they belong to the
same Zoroastrian sphere; today, however, most scholars will agree that the
Middle Persian tradition can hardly ever be taken as a reliable witness to the
meaning or analysis of Avestan words or passages, even in the case of texts that
claim to be translated from Avestan sources. In other words, Middle Persian can-
not be regarded as a continuant of the language of the Avesta, either linguistically
or with respect to the transmission of its contentsall the more so where the Old
Avestan texts are concerned.
The third reason why we are still far from being able to solve all the riddles
posed by the Avestan tradition lies in the circumstances of the transmission of the
Avestan texts themselves. Three essential points must be considered here.
First, we must bear in mind that the entire corpus of Avestan texts was trans-
mitted orally for a long time before the first attempts were made to write them
down. In this respect, the Avestan tradition is very similar to Vedic but again
diametrically opposed to that of Old Persian, the written attestations of which are
practically contemporary with their composition. Oral transmission obviously had
several effects on the shape of the Avestan texts. One such effect, the impact of
which can hardly be estimated today, is that of intentional redaction, which may
have affected the texts several times during the oral period. This effect may be
seen, for example, in the addition of explanatory glosses or the duplication of pre-
verbs and similar elements standing in tmesis, as in Y. 28,10c, where xmaibii
seems to gloss v, both meaning for you:
at .v. <xmaibii.> asn. vad. xvaraiii. vaitii. srauu.
or Y. 48,7a, where the preverbs n down and pait against are duplicated:
3
The figures given here are based on the electronic text editions contained in the TITUS collection,
cp. http://titus.uni-frankfurt.de/texte/texte2.htm.
THE AVESTAN LANGUAGE 167
In metrical passages, these effects can easily be identified, because the added
elements are normally hypermetrical.4
Besides deliberate redactional intervention, there are at least two kinds of un-
intentional factors that may have affected the texts to a certain degree. Although
the Zoroastrian priests no doubt endeavoured to preserve and transmit the original
wording as far as possible unaltered, they could not prevent it from being
influenced over the centuries by what we might style internal and external inter-
ferences. Among the former I should reckon processes of mutual adaptation of
Old Avestan and Young Avestan forms, as in the case of Proto-Iranian word-final
*-ah (< *-as), which, according to Karl Hoffmanns proposal,5 yielded - in Old
Avestan but - in Young Avestan; if this suggestion is correct, we would have to
assume that doublets like vac word, utterance (< *u ah, cf. OInd. vca)
appearing alongside vac in Old Avestan,6 together with other Old Avestan words
ending in - instead of -, reveal Young Avestan influence.
The opposite interference can be seen in cases where Young Avestan shows
an intervocalic voiced stop when we should expect a voiced fricative instead, as
in frte they prosper in Vd. 21,1. The variant frdte with a dental stop is
met with several times in the verse7
yehe. iiaoni. ga. a a. frdte.
by whose deeds the livestock prospers with truth,
In the case of the present root frd- to further, the stop also occurs in other
Young Avestan forms where Old Avestan influence is not so evident. This is true,
for instance, of the various compounds with the participle frdat as their first
member, which appear never to have been written with the fricative.8 But for
these words too an Old Avestan model may be found in frdat .gam, which
occurs once in Y. 33,11.
This raises the question of the evaluation of the so-called Pseudo-Old Avestan
texts. A good example of this genre is the frequently quoted Yeh Htm prayer,
which is normally cited as Y. 27,15:9
4
For the Gathas, a list of preverb duplications is provided in Humbach, Gathas I, 59f.; for the Yasna
Haptahiti cf. Y. 41,1: dadmahic. cmahic. c. <>uuadaiiamah.
5
MSS 22, 1967: 34 (= Aufstze zur Indoiranistik, II, 491).
6
OAv. vac Y. 45,5b; vac Y. 31,20b; cf. YAv. vac Y. 8,4 etc.
7
Y. 19,17; Vr. 2,5; 3,4; G. 4,9.
8
YAv. frdat .fu-, frdat .ga-, frdat .nar-, frdat .vahu-, frdat .vra-, frdat .vspm.hujiiti-,
frdat .xvarnah-.
9
Within the Yasna liturgy proper (cf. below), this prayer is repeated 72 times (Y. 4,26; 5,6; 6,21;
7,27; 10,21; 13,7.8; 17,19; 18,8.9; 19,21; 20,5; 21,1.5; 26,11; 27,15; 28,12; 29,12; 30,12; 31,23;
168 Jost Gippert
32,17; 33,15; 34,16; 35,102; 36,6; 37,5; 38,5; 39,52; 40,4; 41,72; 41,8; 42,6; 43,17; 44,21; 45,12;
46,20; 47,7; 48,13; 49,13; 50,12; 51,23; 53,10; 54,2; 55,7; 56,5; 57,4.6.8.10.12.14.18.20.22.26.29.
32.342; 59,29.33; 60,13; 63,3; 65,18; 67,8; 68,24; 70,7; 71,24.31.
10
Y. 65,16; 69,1; 69,2 (+ rep.); 69,3 (+ 2 reps.). Further quotations of this and the Yeh Htm
prayer are found in the extra passages of the Vispered, Vidvdd, and Vitsp Yat liturgies (cf.
below), esp. within Y. 59,33 and Y. 71,16.
11
An interference of the Gathic strophe Y. 51,22 may be responsible for a certain disagreement in
the manuscript tradition concerning the verse structure of the Yeh Htm prayer; cf. Geldners n.
1 to Y. 4,26. Those manuscripts which insert the Pahlavi translation after pait (Pt4, F2 in Y. 4,26;
in Y. 6,21 also J2, K5, Mf4) may be influenced by the fact that the first verse of Y. 51,22 ends in
pait. Interestingly enough, Mf4, the sister manuscript of Pt4 which was not available to Geldner
when he prepared his edition but which has meanwhile been published in facsimile (K. M.
JamaspAsa and M. Nawabi (ed.), Manuscript D90, Yasn With its Pahlavi Translation, 1-2. Pahlavi
Codices and Iranian Researches, 19-20, Shiraz 1976), inserts the break after vah in Y. 4,26 (cf.
fig. 1, extracted from p. 103 of the facsimile edition, with yehe (!) and vah marked), which
would agree with an original verse structure of 3 11 syllables as proposed by Geldner. (In Pt4,
vah is simply omitted in 4,26.) Mf4 is also important for preserving, though not consistently, the
palatal of vah which was postulated as the original reading by Karl Hoffmann, Festgabe
deutscher Iranisten zur 2500-Jahrfeier Irans, 1971: 69 (= Aufstze zur Indoiranistik, I, 321): within
attestations of the prayer, Mf4 has vah in Y. 7,27, 27,15, and 60,13 (cf. fig. 2, showing Y. 27,15,
extracted from p. 334 of the facsimile edition, with yehe (!) and vah marked). Cf. pp. 180f.
below for further consideration of this question.
THE AVESTAN LANGUAGE 169
instead of the voiced we find in vaa in the clearly related Young Avestan
passage Vr. 16,3.12
yam. n. ahur. mazd. a auua.
yesne. paiti. vah. vaa.
12
The passage is quoted twice in Vr. 24,0 and reappears again in Yt. 13,148.
13
daiti (3 sg. pres. ind. act.) is attested 54 times (Y. 9,22; 65,2; Yt. 2,122; 5,2; 8,49; 10,33. 262. 28.
622. 107; 13,5; Vd. 3,34. 35; 15,2. 3. 9. 11. 13. 15; 18,12. 34. 37; Vr. 7,3; Ny 4,3; N 17. 252. 37. 61.
62. 84. 1052; P. 6(7). 50; Vn. (FrK.) 8. 272. 372. 62. 71-72; Vyt. 5,34; 7,46. 48; Nik. 52. 16. 17. For
daiti there are only two attestations in the Pursinh: P. 23(24) and 34(35); the 3 sg. pres. subj.
act., however, is attested with -- only (daat ): Y. 11,5; Yt. 5,19. 23. 27. 31. 35. 39. 43. 47. 51. 55.
59. 66. 70. 74. 79. 83. 892. 99. 106. 110. 114. 118; 9,5. 11. 15. 19. 23. 27. 32; 11,15; 14,29. 31. 33;
15,4. 9. 13. 21. 25. 29. 33. 37. 41. 434. 442; 16,7. 10. 13; 19,10. 94; Vd. 18,29; 19,92; Vr. 12,4; Ny.
4,8; P. 31(32); Vyt. 4,25. The problem becomes even more complicated if we consider the forms of
the perfect active participle: as against -- in 19 occurrences of the nom.sg. dauu (Yt. 5,17; 10,50.
143; 13,83; 15,2. 4. 9. 13. 17. 25. 29. 33. 37. 41; 19,16; Vd. 2,20. 21; 7,52; 22,7) and 12 of the
acc.sg. dauuhm (Y. 6,1; 16,4. 5. 6; 70,1; Vd. 18,7. 13. 60. 66; S. 2,8. 15. 23), we find -- alone
in the oblique case forms, viz. in 18 attestations of the gen.-abl.sg. dau (Y. 1,1; 4,7; 16,3; 22,1;
24,12; Yt. 13,78. 157; 15,44; Vd. 19,4; Vr. 11,16; S. 1,8.15. 23; A. 3,7. 8. 11; Vyt 7,46. 51) and one
of the (secondary) abl. dauat (Yt. 5,7).
170 Jost Gippert
14
According to Geldners apparatus, the reading vaca is supported by Mf3, Pd, K36, Lb16, Ml2,
while J9 has vae.ca. Another group of manuscripts, including F1, the allegedly oldest Yat
manuscript available, provides quite a different reading: auuat .ca (F1, L11), avat .ca (Jm4),
auuaa.ca (E1, P13, K19), auuadaca (L18, K12, J10, O3), auuaaaca (Pt1). Instead of tat ca
(Mf3, Pd, Ml2), we find dadaca in J9, F1, ..daca in Mb1, dadaci in Jm4, L11, daaca in K19, J10,
P13, E1, L18, K12, and daaca in Pt1. kaica is provided by Mf3, Pd, K36, Lb16, as against
kaeat ca (J9, Jm4), kaeaaca (E1, P13, L18, K19, O3), kaeadaca (Mb1, F1), and kaaaca (Pt1,
J10, K12). It is clear that we have a dichotomy of manuscipts here, Geldners text being based on
the group consisting of Mf3, Pd, K36, and Ml2, all of which are Khorde Avesta manuscripts rather
than Yat manuscripts proper. On the basis of F1 and its descendants, one might prefer to restore
*auuaa ca. *yaa. *kaaca. in just that way and what way so ever, which reminds one of the
curse formula which appears in Yt. 19,57 (ie. ia. yana. ahmi), 60 (ie. ia. yana. ahmi.
auuaea. ia. yana. kahmi.) and 63 (ie. ia. yana. ahmi. auuae. ia. yana. ahmi. auuiia.
ia. yana. ahmi.); but *vaca. *yaa. *kaaca. and I know, in what way so ever also remains
possible (for yaa. kaaca cf. Yt. 4,7; 19,82; Vd. 2,11.15.19, and N. 37). The formula auuaa.
ataa. atahe. iiaonahe. yaa. vati when they blame (somebody) there for this deed here,
which appears five times in Vd. 4,50ff., is too remote to be compared here, albeit it seems to contain
another instance of the development -- > -- (if vati belongs to u id as proposed by
Bartholomae, Airan.Wb., 1322).
15
Handbuch der Orientalistik, 4: Iranistik, 1: Linguistik, Leiden 1958: 8 (= Aufstze zur
Indoiranistik , I, 65); ALF, 97f.
16
Cf., e.g., 3 sg. pres. ind. dahd <YHBWNyt>, Mng- xrad 21,41 (ed. Sanjana, p. 40, l. 6); the
Pzend version, ed. Anti reads <daht > (p. 301, l. 22); 3 pl. pres. ind. dahnd, ibid., 41,16
(<YHBWNd> p. 62, 6 / <dhd> p. 317, 12).
THE AVESTAN LANGUAGE 171
17
E.g. 3 sg. pres. ind. dayd in M 49 I R 3 (Mir.Man. II, 306, l. 13), 3 pres. ind. daynd M 11 V 5
(Waldschmidt-Lentz, Man.Dogm., 557).
18
Thus, e.g., DB 1,95 and 3,32.
19
Thus DB 1,40 and 1,80.
20
Cf. Hoffmann, ALF, 36.
21
Sasan.Arch., 39 ff.
172 Jost Gippert
presume that the whole corpus was not transmitted by the same people, Parthian
influence manifesting itself only in those parts transmitted by Parthians. Alterna-
tively, we might take refuge to the idea of Proto-Parthian influence at the time of
the composition of the texts in question, which again could be interpreted in two
ways: it might mean that there was a dialectal split within Young Avestan as a
spoken language, one dialect showing interference from (neighbouring) (Proto-)
Parthian, or it might mean that the texts were composed by speakers of (Proto-)
Parthian who used Young Avestan as a sacral language. For the latter case, one
might well compare the use of Latin as a sacral language in Europe, which caused
some similar effects when adapted during the past millennium by speakers of lan-
guages as different as French, German, English or Hungarian. Finally, each of the
scenarios envisaged above has a chronological aspect, in that it might be inter-
preted as indicating a certain age for a text containing the feature in question.
We cannot expect to be able to solve this riddle as long as several other
problems raised by the Avestan tradition remain unclarified. If we speak of
parts of the Young Avestan corpus which might have been influenced by Parth-
ian, we presuppose that the Young Avestan texts available to us are in some way
heterogeneous, and we should expect certain features that can be interpreted as
dialectal in the sense outlined above to appear as bundles within these parts. At
present, however, we are still far from being able to establish the existence of
such bundles or to delimit in detail the specific parts of the text which are charac-
terised by them. This unsatisfactory situation is due to several awkward
peculiarities of the Avestan manuscript tradition.
First of all, we have to take into account the fact that most of the manuscripts
which have come down to us contain not separate texts but collections of texts,
the composition of which is not due to their dialectal or chronological uniformity,
not even to their contents, but only to their applicability for certain liturgical
purposes. This is especially true for the Yasna collection, which represents, as is
well known, the texts and text elements to be uttered by the Zoroastrian priests in
the course of the Hama sacrifice. It is for exclusively liturgical reasons that this
collection contains, right in its middle, the oldest Avestan texts available to us,
i.e. the Gathas and the Yasna Haptahiti, as well as such strange Young Avestan
passages as the one following the so-called Hm Yat, which is cited as Y. 11,9:
y.n. auu. at .t. uii. riiidiii.
trahe. madidiii. xuudm. haptdiii.
nauua. dasme. yi.v. yama.
These verses represent nothing but a sequence of quotations from Old Aves-
tan texts which contain the numerals from one to ten, or word-forms resembling
these numerals, partially concealed in unrelated infinitives, which were adapted
to Young Avestan and thus obscured in a cabbalistic way with the result that
there is certainly no point in trying to interpret them meaningfully; cf. the
following synopsis:
THE AVESTAN LANGUAGE 173
are inserted in the position of eight of the ten Vidvdd sections, Vyt. 1 appear-
ing after Y. 27 and Vr. 12,1-5 but before Y. 28. The interrelationship thus estab-
lished can roughly be illustrated in a diagram such as the one given in Table 1.22
Liturgies:
Yasna & Vispered Vidvdd (& Extr.Wg. 1A) Vitsp Yat (& Extr.Wg. 1B)
frauuarne. mazdaiiasn. zarautri. vdauu. ahura.t ka
dti. haa.dti. vdauui. haa.mri. zaini.parti. upairi. gtubii.
huuane.
zarautri. grpti. mri. spti.
a aone. a ahe. rae.
Under these circumstances, it would still be justifiable to treat the four text
collections of Yasna, Vispered, Vidvdd and Vitsp Yat separately, as Wester-
gaard and Geldner did in their editions, were it not for certain deviations in the
arrangement and wording of the Yasna text contained in the extended liturgies.
These deviations, of which Y. 3,24 provides a typical example (cf. Table 2), have
hardly ever been dealt with thoroughly in Avestological literature, and they are
only partially accessible in the editions: while Westergaard listed eight of them,
which he took from Vidvdd Sda manuscripts and K4, under the heading of
Extrakte, Geldner confined his information about such deviations to the critical
22
This table, which agrees with the one provided by Geldner in Grundri II, 11f., shows only the
major deviations. The picture becomes much more complicated when minor divergences are
considered.
THE AVESTAN LANGUAGE 175
Another formula, appearing in Y. 15,4 and Vr. 3,6, mentions instead the fire-
priest, truuax-:
zt : ya. ah. vairii. y. truuax. fr.m. mrt.
rsp: a. ratu. a t ct . haca. fr. a auua. vuu. mraot.
It seems to have escaped notice so far that two further versions of the formula
are met with in the deviant passages of the Yasna contained in the extended
liturgies, within Y. 59,33 and Y. 65,17. Here, two other officiating priests are
named, namely the sraouuarz- and the frabrtar-:
zt : ya. ah. vairii. y. sraouuarz. fr.m. mrt.
rsp: a. ratu. a t ct . haca. fr. a auua. vuu. mraot. (Y. 59,33)
Although both these titles appear several times in other Avestan passages such
as Vd. 5,57, Vr. 3,1, G. 3,5, Vyt. 3(15) or the Nirangistan (67 ff.), it is interesting
to note that it is only from the formulae discussed above that it becomes clear that
the persons bearing them are involved as speakers in the ritual. And it may be
remarked that in both cases the additional attestations thus obtained help to
establish the correct phonetic shape of these titles, which are now confirmed by
the Vispered manuscript K7 of 1288 A.D.supposedly the oldest Avestan manu-
23
H. Brockhaus (ed.), Vendidad Sade. Leipzig 1850 ( repr. Hildesheim 1993).
176 Jost Gippert
script extant. So we can now take it for granted that frabrtar- attendant had
hysterodynamic inflexion with zero-grade root, -br- < *-bhr -, throughout its
Avestan paradigm, as against variae lectiones with -ar- (reminiscent of Vedic
proterodynamic prbhartar-) in Vitsp Yat (K4) and Nirangistn manu-
scripts.24 As for the officer producing obedience, we gain further evidence for
the conspicuous long vowel which appears at the boundary of most compounds
containing the root varz- to produce as their second member, sraouuarz-
agreeing with vstriiuuarz-, huuartuuarz-, duuuartuuarz-, iiaon-
uuarz- and haiiiuuarz- (as against the ambiguous huuarz- /h-u arz-/ and
vohuuarz- /u oh-u arz/).
Whatever one may think about this peculiarity, the deviations from the Yasna
text which occur in the extended liturgies leave us with two dilemmas. The
minor one consists in the problem of how to refer to the passages in question.
Given that Geldner did not bother with them at all, basing his edition on the
Yasna liturgy proper, we have no choice but to confine ourselves to indicating the
enclosing Yasna passages (e.g., Y. 59,33), wherever these can be determined.
Maybe this was one reason why Bartholomae in his Wrterbuch took notice only
of Westergaards Extracts; if he had worked through Brockhauss edition of the
Vendidd Sda, he would have encountered the same problem.
The major dilemma brought about by such deviations is one of the most
striking problems Avestan studies have by now been facing for several centuries,
namely, the question of the evaluation of manuscript tradition. It is true that
Geldners investigations, undertaken in preparation for his edition and continued
during the course of this work, yielded immense progress with respect to the
classification of the manuscript material available to him; and again we must be
grateful to Karl Hoffmann, who significantly improved our understanding of the
matter by reassembling and reformulating the particulars scattered throughout
Geldners Prolegomena. Nevertheless, the knowledge we have gained in this way
does not yet suffice to clarify all the problems connected with the transmission of
the text.
24
Nom.sg. frabarta in HJ, facs.-ed. D. P. Sanjana, fol. 135v, l. 3 (the passage in question is part of
N. 67, p. 110 in Waag's edition); TD, facs.-ed. Harvard Iranian Series 3, fol. 91v, l. 5 has frabarta
at this place as well. In other places, both HJ and TD have frabrt-: HJ 139r, l. 12 / TD 94r, l. 1 (=
N. 68a) and HJ 146v, l. 12 / TD 98v, l. 1 (= N. 71a) (nom.sg. frabrta); HJ 155v, l. 12 / TD 105r, l.
2 (= N. 74a), HJ 157r, l. 11 / TD 106r, l. 2 (= N. 79f) and HJ 159r, l. 5 / TD 107r, l. 7 (= N. 81i)
(gen. sg. frabrtar); HJ 159v, l. 11-12 / TD 107v, l. 6 (= N. 82b) (acc. sg. frabrtrm). In other
contexts, forms containing -bar- are only rarely met with: Geldner notes dat. sg. frabarre instead
of frabrre for Mf2 in Vd. 5,58 (n. 4), frabartrm instead of -brtrm for K11 in Vr. 3,1 and
for K12 in G. 3,5. The assumption of a hysterodynamic frabrtr- further matches with the
MPers. equivalent fraburdr (cf., e.g., the diagram printed in Geldners edition at the beginning of
Vr. 3, which shows the respective positions of the sacrifice participants, among them the fraburdr,
according to the Vispered tradition).
THE AVESTAN LANGUAGE 177
First of all, the information available from Geldners edition is, in the case of
the Yasna tradition, concentrated on the manuscript tradition of the Yasna liturgy
proper, as we have just seen. Although we may well believe that the Yasna part of
the text as contained in Vispered, Vendidd or Vitsp Yat Sda manuscripts is
derived from what we find in Pahlavi Yasna or Sanskrit Yasna codexes, this has
not yet been proved. In other words, it is still open to us to doubt whether we
have to assume that there was a single archetype for each of the major Avestan
texts, as Karl Hoffmann did, or whether there might have been several indepen-
dent first attempts to write them down, depending on their usage in liturgy, which
would have manifested themselves in the divergent manuscript classes estab-
lished by Geldner.
Secondly, the material Geldners edition provides does not suffice for a
thorough reinvestigation of this question. This is not only due to the fact that
Geldner, in preparing his Yasna edition, ignored most of the deviations of the
extended liturgies discussed above, but also to the fact that he regarded several
graphic peculiarities as negligible. This holds true, for example, for the distinc-
tion of <ao> and <a>, <> and < > or <> and <>. With a view to establishing
a critical text, his decision may have been justified, but for investigating the inter-
dependence of manuscripts, especially those which Geldner himself reckoned as
less important, such distinctions may be crucial.
Finally, at least one manuscript that has become available eversince Geldner
finished his Yasna edition is important enough to deserve checking systematically
for the readings it preserves, namely Mf4. It was Geldner himself who belatedly
realised the value of this manuscript, which he proved to be a sister manuscript of
Pt4 and a prominent member of the Iranian Pahlavi Yasna family.25
It is on the basis of these considerations that I decided some years ago to
undertake a project whose objective is the reassembly of the available material in
digital form and the application of new methods based thereupon to the study of
Avestan. As I have described the scope of the project named AUREA in detail
on another occasion recently,26 I shall try to be brief on this subject here.
The starting-point of the AUREA project was the first digitised version of the
Avestan corpus which was prepared by Sonja Gippert-Fritz in the 1980s in
connection with Bernfried Schleraths Avesta Dictionary project in Berlin. As the
electronic version produced at that time contained only the plain text as represen-
ted in Geldners edition, the first task to be undertaken in the present project, with
a view to tackling the problems outlined above, consisted in digitising the data
contained in Geldners critical apparatus. For the Yasna, this task has meanwhile
25
Cf. n. 11 above for a discussion of the value of Mf4.
26
Cf. my paper Indo-iranistisches Textretrieval presented at the conference Indogermanisch und
Indo-Iranisch which has meanwhile been printed in B. Forssman and R. Plath (ed.), Indoarisch,
Iranisch und die Indogermanistik. Arbeitstagung der Indogermanischen Gesellschaft vom 2. bis 5.
Oktober in Erlangen, Wiesbaden 2000, pp. 133-45.
178 Jost Gippert
been finished, thanks to the efforts of Michiel de Vaan, who also collated and
incorporated additional variant readings given in Westergaards edition. As his
second task, de Vaan started recollating several important manuscripts that are
available via facsimile editions or microfilms; besides Mf4, which I have already
discussed, these include the Pahlavi Yasna manuscripts J227 and K5,28 the Vendi-
dd Sda manuscript P129which was the basis for H. Brockhauss edition and
the Vitsp Yat Sda manuscript K4, a microfilm of which was obtained from
the Copenhagen Royal Library. Subsequently, de Vaan collated the readings of
Br2 noted in Geldners Nachla in Marburg University Library, and a microfilm
27
Facsimile edition by L. H. Mills, Oxford 1893.
28
Facsimile edition by K. Barr, Copenhagen 1937.
29
Facsimile edition (produced manually) by E. Burnouf, Paris 1829-1843.
THE AVESTAN LANGUAGE 179
30
Facsimile edition by K. Barr, Copenhagen 1944.
180 Jost Gippert
As was stated above, however, the primary goal of our work on the manuscript
tradition is to establish a reliable database of variant readings which can enable us
to investigate the consistency of readings and the interdependency of manuscript
classes. One example may suffice to show how such a database can be used, what
has been achieved and what it still to be desired.
In connection with the Yeh Htm prayer (cf. p. 168 above), I mentioned the
form vah better which distinguishes the Pseudo-Old Avestan text from its
presumptive model, Y. 51,22, with Gathic Old Avestan vahii. In Geldners
edition, the younger form is regularly written vah, with a nonpalatal . Leaving
aside the occurrences in the Pseudo-Old Avestan Yeh htm prayer, we find
five attestations of this comparative form in the Young Avestan parts of the
Yasna, in Y. 59,30.31, 65,11.14 and 71,13. For these five occurrences, a total of
93 manuscript readings is now known either on the basis of the information given
by Geldner or through the collation of manuscripts undertaken as part of the pre-
THE AVESTAN LANGUAGE 181
sent project. Fourteen of these readings show the palatalised velar nasal , which,
according to Karl Hoffmanns reasoning (cf. n. 11 above), must be taken as the
original variant. By listing them according to the manuscripts, we find Geldners
assumptions about the close relationship of Mf4 with Pt4 and Mf1 confirmed;
moreover, the value of the branch they represent, i.e. the Iranian Pahlavi Yasna,
stands out clearly, all the more so since K5, which represents the Indian branch of
the Pahlavi Yasna, agrees with them in at least one case, Y. 65,14. Surprisingly
enough, Pt4 and Mf4 have vah with a plain velar nasal in Y. 71,13, where K5
agrees with Mf1 in reading vah again. In other manuscripts, vah appears
only twice (K36, Pd). The findings may be summarised in tabular form as
follows:
While all this agrees with what we might expect, two questions remain open.
First, there is no way yet to evaluate the two other manuscripts reading vah
in Y. 65, namely K36 and Pd. These two manuscripts were classified by Geldner
as belonging to the Khorde Avesta type, which means that they cannot be
assigned a position within the stemmata of Yasna manuscripts off-hand. Never-
theless, it is worth while trying to find out whether or not the readings they pro-
vide in Yasna passages depend on any particular branch of the manuscript tradi-
tion, a task which Geldner did not even attempt. For this purpose it is necessary to
study in greater detail the transmission of lesser liturgical collections such as the
so-called May Yat, within which Y. 65 is enclosed in the manuscripts in
question. As a matter of fact, neither Geldners nor any other edition of the
Avesta gives any information about the elements that constitute this liturgy31, the
editorial practice having always been determined exclusively by greater units
such as the Yasna collection or the Vidvdd.
31
For the May Yat or May Zhr, cf. Geldner, Grundri, II, 9 nn. 10 and 11, according to
whom the two titles are mentioned in the manuscripts Mf3 and (Wilsons Ms. Nu. 1 =) W1,
respectively; according to the Prolegomena to his edition (p. xi), the May Yat includes Y. 65 in
Mf3.
182 Jost Gippert
Secondly, it is not at all clear whether the list of readings Geldner provides is
complete in the sense that all the manuscripts he had to hand were checked in all
five cases and whether the variant reading with palatal is really restricted to the
few manuscripts named. Today, all we can do is to try to reconstruct Geldners
practice in this respect by comparing the information he gives for manuscripts
such as J2 or K5 with what a recollation of these manuscripts may reveal. One
example may suffice to show that even in the case of these manuscripts, which
Geldner valued extremely highly, he neglected a significant number of variants.
For the relative pronoun yeh, which introduces the Yeh Htm prayer, Geld-
ner mentions three cases where J2 has a short final e, thus yielding a normal
Young Avestan yehe instead of the expected Pseudo-Old Avestan form: Y. 21,1
(in the prayer formula itself), Y. 61,1 (where the introductory words are used to
build a hypostatic feminine noun denoting the prayer, yeh.htmca.
hufriiatm. fra iimah) and Y. 12,1 (in another Pseudo-Old Avestan text,
the nism. dauu prayer, in the formula yeh. gu). He gives no indication,
however, of the fact that yehe is by far the most usual variant in J2, appearing in
as many as 62 further recitations of the Yeh Htm,32 one further attestation of
the name of the prayer (Y. 72,1) and nine other Pseudo-Old Avestan passages
within the Yasna.33 As a matter of fact, the correct form yeh is attested only
once in J2, at the point where Y. 15,2 (itself an adaptation of Y. 51,22, cf. above)
is repeated in Y. 69,2, a piece of information which may indeed be significant, all
the more so since the two other repetitions of the same formula in Y. 69 (1 and 3)
show the usual yehe. In the light of these observations, we may expect that the
recollation of less-studied manuscripts such as J3 or Mf1 will reveal a great deal
of additional information.
It goes without saying that the database we have been establishing since 1996
now can and will also be used as a basis for up-to-date editions of the texts.
Unlike Geldners, these editions can be so arranged as to meet the requirement of
illustrating the position of a given passage within its liturgical context, not only
by distinguishing the different liturgies, as discussed above, but also by con-
trasting the divergent types of text that constitute them: coherent hymnic or poetic
structures such as the Gathas or the Hm Yat; formulaic prayers such as the
Ahuna Vairiia and their repetitions; litanies such as those which abound in the
Young Avestan parts of the Yasna and in the Vispered; quotations and adap-
tations of strophes or verses from both Old Avestan and non-Old Avestan texts;
and simple allusions to elements contained in other contexts. It is conceivable that
32
Y. 5,6; 6,21; 7,27; 10,21; 13,7.8; 17,19; 18,8.9; 19,21; 20,5; 21,5; 26,11; 27,15; 28,12; 29,12;
30,12; 31,23; 32,17; 33,15; 34,16; 35,10; 37,5; 38,5; 39,5; 40,4; 41,7; 43,17; 44,21; 45,12; 46,20;
47,7; 48,13; 49,13; 50,12; 51,23; 53,10; 54,2; 55,7; 56,5; 57,4.8.10.12.14.18.20.22.26.29.32.34;
59,29.33; 60,13; 63,3; 68,24; 70,7; 71,24.31. In Y. 36,6 and Y. 67,8, the word is damaged but its
final -e is clearly legible. In Y. 4,26 and 42,6 the manuscript is damaged precisely at the end of the
word; in Y. 41,8 and Y. 57,6 the prayer is omitted in J2.
33
Y. 12,13; 15,2; 58,4; 63,1; 65,16; 69,1.3.
THE AVESTAN LANGUAGE 183
34
RV 1,26,2.10; 30,4; 54,3; 57,4; 75,1; 78,5; 83,3; 84,19; 91,10; 93,2; 94,8; 101,1; 114,6; 144,7;
145,2; 2,31(222),5; 3,10(244),5; 3,33(267),8; 4,33(329),5; 5,1(355),12; 11(365),5; 22(376),4;
39(393),5; 54(408),15; 6,48(489),11; 7,8(524),6; 96(612),1; 101(617),5; 8,8(628),11; 19(639),12;
24(644),20; 25(645),20; 39(659),2; 43(663),27; 46(666),14; 61(670),1.9; 66(675),5; 74(683),1;
101(710),5; 9,101(813),13; 103(815),1; 10,17(843),14; 37(863),6; 50(876),6; 64(890),10;
97(923),14; 108(934),8; 122(948),2; 150(976),2.
35
Old Avestan attestations: Y. 31,22; 32,5; 33,2; 34,1.
36
RV 1,76,4; 2,18(209),3; 14(248),6; 4,1(297),15; 5(301),14; 25(321),2; 5,29(383),6; 6,62(503),5;
10,151(977),1.
37
vacah-c, Y. 47,1 (and repetitions); 48,4; vcas-ca, R V 2,32(223),3.
184 Jost Gippert
While varc might still be, as a regular nom.-acc. pl. of *varcah- might,
energy, the equivalent of Ved. vrcsi,43 hc would, as single word, have to be
interpreted as the instr. sg. of a verbal root noun hic- meaning the act of
pouring; this solution was, hesitantly, envisaged by H. Humbach44 and further
supported by J. Kellens.45 Another solution was put forth by S. Insler who
proposed to see here a locative sg. *varcah-ca,46 which would be equivalent to
Ved. vrcasi(-ca).47
It goes without saying that all the solutions mentioned are quite acceptable
from a syntactic point of view, in that either an acc. pl. *varchi-ca or an acc. pl.
38
Post-Vedic vacasi, Mbh. 12,59,138; vacasm, Mbh. 14,44,5.
39
Y. 31,1; 33,8; 35,9; 58,6.
40
R V 1,145,3; 3,33(267),10; 4,3(299),16; 38(334),10; 6,32(473),1; 52(493),14; 7,23(539),1;
8,101(710),7; 10,65(891),13; 66(892),11; 95(921),1; 108(934),6.
41
Airan.Wb., 1367.
42
The manuscripts reading varc are, according to Geldner, K5, Pt4, K4 and Lb2, to which Mf4
and Br2 can now be added.
43
Thus AVS 4,22,3; unaccented varcsi twice in AVP (3,21,2, 6,19,9).
44
Gathas II, 88 (cf. already his earlier edition Die Gathas des Zarathustra, Heidelberg 1959; II, 37).
45
Les nom-racines, 88 sq.; Kellens-Pirart, II, 328.
46
S. Insler, The Gths of Zarathustra, TehranLige, 1975, ad loc.
47
Cf. unaccented varcasi in AVP 20,53,7.
THE AVESTAN LANGUAGE 185
varc might depend from n ... dadat they lay down, in concord with the acc.
pl. xrat, and a loc. sg. *varcah-ca might also depend from the same verb, in
concord with the loc. sg. .hii in the tying; while hc, if taken as the instr.
sg. of a verbal noun pouring, might represent a syntagmatic unit together with
the following fradiuu, taken as an adjective meaning everlasting.
The question posed above of the regularity of sound correspondences is in-
volved in this problematic case both internally, i.e. with respect to the representa-
tion of Old Avestan sounds by graphemes in the manuscripts, and externally, i.e.
with respect to the relationship of the proposed readings to their Vedic counter-
parts. The former would present us with the problem of evaluating the difference
between the variant readings, i.e. - vs. -, in connection with the divergent
interpretations. As for the latter, we would have to investigate the question
whether we are right at all in assuming *varchi to be the regular outcome of the
Proto-Indo-Iranian preform which yielded Vedic vrcsi. Within the AUREA
project, several steps have already been undertaken with a view to these investi-
gations, the preparation of the database of variant readings being only one of
them. Another preparatory step consisted in the compilation of a list of acknow-
ledged or alleged equivalences between Old Avestan and Vedic on the level of
word-forms, for the purpose of analysing with a special computer programme the
relative chronology and the consistency of the sound correspondences involved. It
may be of interest to note that the list of presumable equivalences contains about
900 Old Avestan word-forms out of a total of less than 2000, which indicates a
very high degree of agreement indeed. The computer programme (developed by
R. Gehrke and K. U. Bux) is still in the state of testing; we hope to be able to
present it (and the results it offers as to the relationship of Old Avestan to Vedic)
to the public within a few months.
In the case of problems such as that of varc.hc in Y. 32,14, however, it is
not only questions of sound correspondences that must be considered. As a matter
of fact, a comparison with Old Indic may be helpful on many other levels as well,
both linguistic and philological. As in many other cases, it might be worthwhile
investigating the textual usage of Ved. vrcas- and the other equivalents of the
Avestan lemmata that occur in the passage in question, a method well established
in Avestan studies. If we do so, we will soon come across two verses of the
Atharvaveda Sahit (AVS 4,8,5-6), where vrcas- occurs in a narrow syntactic
juncture with the verb sic- to pour, thus immediately reminding us of the con-
spicuous hc of Y. 32,14:
The waters of heaven that revel with milk, in the atmosphere or also on the earth,
with the splendor of all those waters do I pour upon thee.
The heavenly waters, rich in milk, have poured upon thee with splendor;
that thou be an increaser of friends, so shall Savitar make thee.
Even though krtu- is not used in a plural form in this mantra, its co-occur-
rence with vrcsi is striking enough to be taken as a parallel to the Avestan
verse in question. It may be added that in the atapatha-Brhmaa, where the
usage of the given mantra within the Vjapeya ritual is discussed (BM 5,2,1,15),
the four terms are resumed in the sequence yas-, ndriya- and vry-, arranged
syntactically as direct objects of datte he lays (into himself):
As a grhma- in (close) connection with this one, even the kavis (continue to) lay
down (on him) power[s] and splendour[s], by daily pouring (?), whenever they get
near the deceitful one to assist (him), and whenever the cow is ordered to be killed, to
assist (him) who inflames the one who is hard to burn.
My aim in taking up the example of Y. 32,14 has been to illustrate why I
believe that further progress in Avestan studies can only be achieved by taking a
closer look at the similarities and interdependencies which connect the Avestan
language with Vedic Old Indic. For studying Vedic, it may not be necessary to
compare Avestan; but for the study of the Avestan corpus, the evidence of Vedic
should never be underestimated.
Jost
Digital unterschrieben von Jost Gippert
DN: cn=Jost Gippert, o=Universitt
Frankfurt, ou=Vergleichende
Sprachwissenschaft,
Gippert email=gippert@em.uni-frankfurt.de,
c=DE
Datum: 2011.12.28 01:28:10 +01'00'